Statistically Speaking...
News and views from the Statistics Department
Official Newsletter of the Department of Statistics, University of Warwick
Issue 5: January 2015
News and Events
Welcome to the fi h issue of
“Sta s cally Speaking...” - a
publica on designed for current
and past students and staff of the
Sta s cs Department at the
University of Warwick.
The results of the 2014 REF (Research Excellence Framework) exercise have now been published, and Warwick
Sta s cs and the Warwick Mathema cs Ins tute together
have been ranked 3rd in the UK with more than 90% of our
research ac vity assessed as either interna onally excellent or world leading.
Thank you to all who contributed
to this issue or par cipated in its
produc on in any way.
Our research environment at Warwick is rated by REF 2014 as the very best in the UK for mathema cal sciences (achieving the maximum possible score of 100% at 4*).
Vassili
Ingram
The University of Warwick as a whole is now ranked (among universi es) 7th in the UK for our research, with 14 academic departments ranked in the UK top ten for their disciplines.
Kolokoltsov,
Eleanor
RSS working party on REF and league tables
Professor David Firth is part of an ad hoc working group formed in early 2015 by the Royal Sta s cal Society to examine the use of REF
results in published research rankings and more general "league tables" of universi es/departments.
Tom Nichols has received a 2 year grant from Warwick's Brazil Partnership Fund to support con nued collabora ons with recent-PhD Lilia
Costa. Now that Lilia has return to her home in Brazil, this grant is vital for funding return her visits to Warwick. It will also fund visits for
Tom & Jim Smith to Brazil to help organise a conference in San Paolo on Network Science.
Dr Heather Turner, sta s cal consultant and currently an Associate Fellow of Warwick Sta s cs, has been invited to give an intensive oneday course Introduc on to Generalized Nonlinear Models in Zurich on 11 May 2015. The course is based on Heather's work as a Senior Research Fellow at Warwick, which included the development of award-winning so ware package gnm.
Job Opportunities for Female Postdocs
Applica ons for the 2015 L’Oréal-UNESCO UK & Ireland For Women in Science Fellowships open 2nd February 2015
L’Oréal UK & Ireland, the UK Na onal Commission for UNESCO and the Irish Na onal Commission for UNESCO,
with the support of the Royal Society have partnered together to provide a dedicated UK & Ireland For Women in Science Fellowship programme for women scien sts at post-doctoral level to enable and/or facilitate
promising scien fic research in the life and physical sciences.
This year five Fellowships will be awarded to outstanding female postdoctoral scien sts to assist them with
their research. The Fellowships, each worth £15,000 (equivalent Euros for candidates in Ireland), are tenable at any UK or Irish university or
research ins tute to support a 12-month period of research and for the first year we will be including maths, engineering and computer science.
The Fellowship money can be spent in any number of innova ve ways to enable women scien sts to pursue their careers and facilitate
world class research - such as buying equipment, paying for childcare, or funding travel costs to an overseas conference.
Key Dates:
Applica ons open – Monday, 2nd February Applica ons close – Friday, 13th March
Shortlist published – Tuesday, 2nd June
Awards ceremony – Tuesday, 23rd June
Visit www.womeninscience.co.uk to apply
http://www2.warwick.ac.uk/fac/sci/statistics
KAUST Undergraduate Poster Competition 2015
The Sta s cs department would like to extend their congratula ons to Daniel Wilson-Nunn, a third-year student
on the Mathema cs and Sta s cs degree programme, for his selec on for the King Abdullah University of Science
and Technology (KAUST) Undergraduate Poster Compe on 2015.
TRAVEL REPORT
WOLFRAM SCIENCE SUMMER SCHOOL 2014
“THE ROUGHNESS AND FRACTAL BEHAVIOUR OF BITCOIN COMPARED TOOTHERMARKETS”
Between 30th June and 18th July 2014, I participated in the Wolfram Science Summer School held at Bentley University, Boston, Massachusetts, USA. The Summer School, organised by Wolfram Science offers a group of international students from varying backgrounds the opportunity to produce a piece of research using the ideologies from Stephen Wolfram’s “A NEW KIND OF SCIENCE” and the software package “WOLFRAM
MATHEMATICA”. My aim was to implement techniques in investigating randomness in financial products and examine methods of modelling
financial products.
The majority of the Summer School consisted of lectures given by various members of the Wolfram Research community on various areas from enumeration of polynomials to image manipulation in MATHEMATICA. In addition to these lectures, “Live Experiments”1 were performed by
Stephen Wolfram (founder of Wolfram Research and the Summer School), where an unsolved problem in Mathematics or Computer Science was examined using MATHEMATICA in front of the students, with input from the students as well as staff and instructors from Wolfram
Research. Advanced MATHEMATICA training was given by senior programmers from Wolfram Research.
After meeting with Stephen Wolfram, Todd Rowland (academic head) and Hector Zenil2, my instructor for the duration of the School, I
decided that my project would involve visualising transactions of Bitcoin. After working on this, my project evolved to encompass studying
the price history of Bitcoin and comparing it to precious metals, stock indices and exchange rates.
In studying the historical prices of Bitcoin, I utilised a number of techniques from the fields of Statistics, Finance, Computer Science and Mathematics

The correlation between movements of prices

Analysis of the logarithm of daily returns and comparing to the normal distribution

Techniques from Information Theory such as Entropy, Compression Ratio and Approximate Entropy (ApEn)

Fractal dimension using Box-Counting Estimator3 and the Hall-Wood Estimator4
Learning to use these different techniques allowed me to gain an insight into many different disciplines and has opened up new areas of interest for myself, namely fractal geometry and fractal analysis of financial investment products5.
I am continuing with my project so as to author a paper with my instructor, Hector Zenil, with the above title. The results have led to a number of
interesting conclusions regarding Bitcoin and its similarities to other investment products:
Correlation between movements of Bitcoin prices between 17/07/2010 & 03/07/2014 and gold prices between 03/01/1900 & 07/07/2014
when scaled to the same time period is 0.946. The same with silver prices gives a value of 0.907.
The logarithm of the daily returns of Bitcoin, gold and silver all follow a similar distribution which is very dissimilar to the normal distribution.
This is in contract to the logarithm of the daily returns of the GBP/USD, CHF/USD and EUR/USD exchange rates, as well as the FTSE100 and
the NASDAQ.
The fractal dimension - calculated using the Hall-Wood Estimator - of the price history of all 5 afore mentioned products when compared to
both Bitcoin and Litecoin again grouped precious metals with cryptocurrencies.
1
http://www.wolfram.com/broadcast/video.php?c=153&v=791
2
http://www.hectorzenil.com/
3
Gneiting, T., Sevˇc´ıková, H. and Percival, D. B. (2012). Estimators of Fractal Dimension: Assessing the Roughness of Time Series and Spatial Data. Statistical Science , No.2, 247 – 277.
4
Hall, P. and Wood, A. (1993). On the performance of boxcounting estimators of fractal dimension. Biometrika
5
Mandelbrot, B. B. (1982). The Fractal Geometry of Nature. W. H. Freeman and Co., San Francisco, CA.
http://www2.warwick.ac.uk/fac/sci/statistics
246 – 252.
Warwick encounters at Joint Statistical Meeting, Boston
An anomalously high concentra on of Warwick Sta s cs PhD
Alumni encountered on a sunny a ernoon in Boston. (Flavio
Goncalves, Mouna Akacha, Piotr Zwiernik)
Keegan Kang - now in the second year of a Sta s cs PhD at
Cornell, working on Big Data.
Summer Research Project: Mass Transportation and Mathematical Finance
In July 2013, I undertook a project with Professor David Hobson to study a specific topic in Mathema cal Finance: robust arbitrage-free pricing of forward-start op ons. Specifically, we studied the following ques on: given probability distribu ons for the start value and terminal
value of the underlying asset of a forward -start op on, what are the lowest and highest prices of the forward-start op on that do not admit
an arbitrage (risk-free profit) in the market?
These highest and lowest prices depend only on the probability distribu ons of the value of the underlying asset on the start and terminal
dates of the forward-start op on; they do not require any assump ons on the movement of the asset value between these two dates (these
prices are known as robust or model-independent prices). In theory, the probability distribu ons of the asset values on the two dates may
be inferred from current prices of vanilla call and put op on prices.
Iden fying the robust prices involves solving a linear programming problem that is similar to a transporta on problem. The task is to find
ways of transpor ng probability mass from the star ng distribu on to the terminal distribu on of asset prices that maximise or minimise
the expected payoff of the forward-start op on, subject to the constraint that the asset price process must be a mar ngale (i.e. the expected price of the asset at any future me must equal its current price).
I developed a MATLAB applica on that provides an approximate solu on to the linear programming problem. The applica on reports the
prices and provides a visualisa on of the joint distribu ons of asset values on the start and terminal dates that produce the maximum and
minimum prices. The applica on was used to study characteris cs of the solu ons of the problem, and it is intended that this work will enable the iden fica on of analy cal solu ons for robust prices.
Edmund Wall
What’s on
Other Statistics Seminars:
Big Data in Cancer
CRiSM Seminars
Seminars will be held in B1.01 (Maths)
Wednesday, 18th March
MS.01
Zeeman Building
Friday, 6 February
14:00—16:00
Leonard Held (ETH Zurich), Gareth Peters (UCL)
WDSI Workshops
For more information, including the
registration form, visit:
http://www2.warwick.ac.uk/fac/sci/wdsi/event
s/yobd/cancer/
Friday, 20 February
14:00—16:00
Li Su (Cambridge), Marina Knight (York)
http://www2.warwick.ac.uk/fac/sci/statistics
SF@W
Wednesday, 11 March (C0.08)
14:00—15:00
Tim Leung (Columbia)
RSS West Midlands Local Group Meetings
More information can be found via the
group’s website
https://sites.google.com/site/rsswmlg/forthco
ming-meetings
Young Researchers’ Meeting
Meetings take place on Tuesdays in C0.06
Estimating Genetic Heritability
Tom Nichols has been published in Proceedings of the Na onal Academy of Sciences. Based on work with former Warwick student Tian
Ge, the work provides an accelerated method to es mate gene c heritability with unrelated subjects. The new method is several orders
of magnitude faster than exis ng methods and is suitable for use with high-dimensional data like MR images of the brain. Ref:
Tian Ge, Thomas E. Nichols, Phil H. Lee, Avram J. Holmes, Joshua L. Roffman, Randy L. Buckner, Mert R. Sabuncu, and Jordan W.
Smoller. Massively Expedited Genome-wide Heritability Analysis (MEGHA). Proceedings of the Na onal Academy of Sciences of the United States of America, in press.
Figure: These are the brain regions iden fied with sta s cally significant heritability of cor cal thickness. This analysis, es ma ng heritability with unrelated individuals and compu ng P-values with resampling methods, is only prac cal due to the accelerated inference
methods proposed in the paper.
Conveying geographical uncertainty
It is always difficult to communicate your uncertainty about something. Psychologists have long shown that mispercep ons and misunderstandings abound when we talk about the likelihood of something happening, either qualita vely or probabilis cally. Difficult though it is,
we do have to communicate about and deal with uncertain es. This is par cularly true in crises and rapidly evolving events. RISCU is working on a contract to help COBR, government's crisis management group deal with geographical uncertainty in the event of a radia on accident. Our brief is to find ways of providing assessments of the likely spread of radia on in the event of an accident at a nuclear power
plant.
Geographical uncertainty is par cularly difficult to convey because it is so easy to confuse contours of the atmospheric plume of radionuclides with contours of probability on a map. We have to find a more ac on-oriented, less confusing way of presen ng the informa on. Add to that the fact that government ministers are thankfully very unfamiliar with handling radia on accidents, and you begin to
appreciate that the challenges that our project brings.
Simon French
Contributions to the next edition are most welcome and should be sent to statspeak@warwick.ac.uk